Skip to content
Probabilistic LSTM language model to generate Harry Potter texts. Implemented in TensorFlow
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


Word-level LSTM to train language models. The data directory provides the concatenatation of all seven Harry Potter books in a text file as input, but any piece of text can be used for the basis of the model.

To use, clone the directory, run to train the model, and to generate arbitrarily long text samples.

The code also provides the option of using pretrained word embeddings such as word2vec or GloVe.

You can’t perform that action at this time.